STRAIGHT-LINE CORRESPONDENCES FROM AN IMAGE SEQUENCE FOR AUTONOMOUS VEHICLE DRIVING
Gian Luca Foresti, Carlo S. Regazzoni · 1993
In the guidance system for an autonomous vehicle, the vision task is particularly important, as it is fundamental for the system to be able to recover the structure of the surrounding environment, to recognize other vehicles present in a scene, and to estimate their motion. For the algorithm presented in this paper, a token-based approach to the matching process is applied. This uses straight lines as significant features. Straight lines are extracted by means of a modified version of the Hough Transform, called the Direct Hough Transform (DHT), which includes the information about the endpoints of segments. The algorithm proposed for the matching of straight lines is particularly fast, as it needs only three steps. Firstly, the search range is reduced by working directly in the Hough parameter space. A matching function for each pair of segments selected previously is then determined. The segment pairs that satisfy the match are then extracted. The algorithm has been tested on both synthetic and real images representing road scenes, and good results have been obtained. (A) For the covering abstract see IRRD 859610.